In the OSL dating of sediment, the scatter in equivalent dose (D e) between grains is almost always larger than would be expected due to counting statistics alone. Some scatter may be caused by insufficient (partial) bleaching of some of the grains prior to deposition. In order to date partially bleached sediment, it is essential to estimate the amount of scatter caused by other processes (e.g. grain-to-grain variability in the natural dose rate). Measurements of such scatter are performed at the single-grain level; by contrast, most OSL dating is performed on multi-grain subsamples, for which grain-to-grain scatter is reduced through averaging. Here we provide a model for estimating the expected scatter (i.e. excluding that caused by partial bleaching) for multi-grain aliquots. The model requires as input the single-grain sensitivity distribution, the number of grains in the sub-samples, and the expected scatter at the single-grain level, all of which can be estimated to an adequate degree. The model compares well with measured values of scatter in D e, determined using aliquots of various sizes, and can be used to help produce a minimum-age D e from multi-grain subsamples that is consistent with single-grain data.